Generalized LM-inverse of a matrix augmented by a column vector
Introduction
Let us begin by considering a set of linear equationswhere B is an m by n matrix, b is an m-vector, and x is an n-vector.
The generalized LM-inverse of the matrix B is the matrix such that the solutionminimizes bothandwhere L is an m by m symmetric positive definite matrix and M is an n by n symmetric positive definite matrix.
Below are the four conditions for the generalized LM-inverse [1].We note that the generalized LM-inverse is the more general kind of the Moore–Penrose inverse. The concept of Moore–Penrose (MP) inverses was first introduced by Moore [2] in 1920 and later independently by Penrose [3] in 1955. In 1960, Greville [4] gave the first formulae for recursively determining the Moore–Penrose inverse of a matrix. His algorithm provides an update of the MP inverse of a matrix whenever new information becomes available. As a result, the recursive formulae have found extensive use in many areas of applications. Among them are statistical inference [5], filtering theory, estimation theory [6], system identification [7], optimization and control, and most recently analytical dynamics [8], [9]. In 1997 Udwadia and Kalaba [10] provided an alternative and simple constructive proof of Greville’s formulae, and later [11], [12] developed recursive relations for different types of generalized inverses of a matrix including the least-squares generalized inverse, the minimum-norm generalized inverse, and the Moore–Penrose (MP) inverse of a matrix.
Recently, the recursive formulae for the generalized M-inverse [13], [14] and for the generalized LM-inverse were obtained. Those for the generalized LM-inverse were proved constructively [15]. In this paper, we provide a much simpler and alternative proof for the recursive formulae of the generalized LM-inverse, , of any given matrix, B, partitioned as , where A is an m by n − 1 matrix and a is a column vector of m components. We show that the four conditions of the generalized LM-inverse of the recursive formulae are satisfied. Besides its inherent simplicity, our proof requires several subsidiary properties of the generalized LM-inverse of a matrix, many of which appear to be hereto unknown; they are presented in the Appendix. More general than the generalized M-inverse, the generalized LM-inverse finds use in an even wider range of application areas than the Moore–Penrose inverse – areas ranging from system theory, statistics, filtering, control theory, and optimization, to signal processing and mechanics.
Section snippets
Result
For any given matrixits generalized LM-inverse formulae are given bywhere A is an m by (n − 1) matrix, a is a column vector of m components, , , β = qTMq, , , and . Note that L is a symmetric positive definite m by m matrix, andwhere M is a symmetric positive definite n by n matrix, M− is the symmetric positive definite (n
Conclusions
The recursive formulae for obtaining the generalized LM-inverse of any general matrix augmented by a column vector were first given in Ref. 15. There they were derived in a constructive manner. We herein provide an alternative proof of their formulae by directly verifying that the four conditions of the generalized LM-inverse are satisfied, thereby confirming the validity of the formulae, and providing several new auxiliary results related to these generalized inverses.
References (15)
Least-squares solution of equations of motion under inconsistent constraints
Linear Algebra and Its Applications
(1995)- et al.
A unified approach for the recursive determination of generalized inverses
Computers and Mathematics with Applications
(1999) - et al.
An alternative proof for the recursive formulae for computing the Moore–Penrose M-inverse of a matrix
Applied Mathematics and Computation
(2006) - et al.
Generalized Inverses of Matrices and Its Applications
(1971) On the reciprocal of the general algebraic matrix
Bulletin of the American Mathematical Society
(1920)Generalized inverse for matrices
Proceedings of the Cambridge Philosophical Society
(1955)Some applications of the pseudoinverse of a matrix
SIAM Review
(1960)
Cited by (4)
About the generalized LM-inverse and the weighted Moore-Penrose inverse
2010, Applied Mathematics and ComputationAn inverse dynamics approach based on the fundamental equations of constrained motion and on the theory of optimal control
2020, Lecture Notes in Mechanical EngineeringOn the Use of the Udwadia-Kalaba Equations for the Nonlinear Control of a Generalized Van Der Pol-Duffing Oscillator
2020, Lecture Notes in Networks and SystemsA Meta-Cognitive Learning Algorithm for an Extreme Learning Machine Classifier
2014, Cognitive Computation